Introduction

This portfolio aims to analyze a specific branch of hip-hop, namely industrial hip-hop. Industrial hip-hop, in my opinion, is one of the most interesting and bizarre genres in rap music. Being a fusion of industrial music and hip-hop, the genre showcases artists blending harsh, mechanical, transgressive or provocative sounds accompanied with rap and rhythm. Due to it being a niche genre, one which many people may not be familiar with, I believe it would be a very interesting topic to explore in this online portfolio. The goal of this portfolio is to research how Spotify interprets and understands industrial hip-hop by analyzing features and songs in the genre to see how this branch differs from other Hip-Hop genres.

My corpus consists of my own playlist titled “Industrial hip-hop”, which contains 40 songs of the genre. The playlist contains many songs in the genre, particularly from artists such as JPEGMAFIA, clipping and Death Grips, as these three are among the most well-known artists in the Industrial Hip-Hop genre.


Secondary link to playlist

How energetic and loud is industrial hip-hop?


This visualization shows the relations between loudness and energy in Industrial music. The tracks were taken from the industrial hip-hop playlist. Size is determined by the track’s popularity while color is determined by the track’s valence. This plot shows a clear correlation between loudness and energy among these artist, namely the louder the songs, the more energetic they are. With this we can get a better understanding of what makes a hip-hop song industrial. In terms of valence and popularity there doesn’t seem to be an obvious pattern.

A couple of chromagrams


The first plot shows the chromagram of “I’ve seen Footage” by Death Grips. The somewhat noisy nature of industrial hip-hop makes it very difficult for spotify to recognize and distinguish certain chords. This is very apparent in this song as there are very few major contrasts in magnitude. Most of the song seems to quite consistently switches between C and F#, with a slight change between 163-182s. The switch between C and F# is also known as a tritone, or the devil’s interval. These intervals are extremely dissonant yet they frequently appear in other Industrial hip-hop tracks.

The second plot analyses “Ain’t it Funny” by Danny brown, which switches intervals between G and C#. G is used during the multiple choruses, while C# is used for the many verses where the artist then raps. This again is another False Tritone, similar to “I’ve seen Footage”.

The third chromagram analyses “Silver Rain Fell” by Scorn, an artist from the 90’s which has had an influence on the development of Industrial music and possibly Industrial Hip-Hop. This track mainly switches between G/Gb and E/Eb, along with C#. The lines represent changes in the main verse, which seem to be very minor.

What is the structure of “I’ve seen footage”?


This page shows the self-similarity matrices of the track “I’ve Seen Footage” by Death Grips. This song is structurally quite interesting since we don’t really see the typical chorus-verse structure throughout the entire track The song builds upon the chorus-verse structure with 3 brief sections at the start, visualized by the three small squares up until around the 30s mark. From there it takes a more familiar form up until around 120s when it calls back to a section from the intro. Afterwards we get a longer than usual verse followed by a bridge at around 160s transitioning into the chorus.

Taking a look at keys and chords


This page shows the keygram and chordgrams of two songs: “HAZARD DUTY PAY!” by JPEGMAFIA, and “Say the Name” by clipping. These keygrams and chordgrams show the matching of each key/chord per time frame, with lighter colors being more often matched and darker colors less often.

With JPEGMAFIA’s track we can see that spotify cannot identify which keys are used in the track, resulting in a completely yellow keygram. The chordgram shows that the song seems to consist only of 7th chords, but appearances can be deceiving. We used the euclidean distance metric to match chords to points in the songs. However, Spotify seems to think that every pitch class is active through the entire song. Seventh chords will always match more strongly than chord triads in this case, simply because they consist of more notes.

Clipping’s keys seem to be in C# mainly, but the mode is rather ambiguous. Its chord progression seems to chiefly consist of the I, II and VI chords, I think that I -> II -> VI progression is most likely.

Struggling to find the beat


Let’s take a look at another track from Death Grips, namely “Have a sad ### BB” and generate a few tempograms. I chose to analyze this song seeing as it uses some rather bizarre mixing and sampling throughout the entire track, which can make it rather difficult to find the correct tempo, and see whether Spotify could then pick up on the correct tempo. The measured tempo in the top visualization at first glance looks incredibly messy and all over the place, though a keen eye might see that there does seem to be a couple of vague lines at around 300BPM and 600BPM. The perceived tempo in the bottom plot also look chaotic, however there is a much clearer tempo visible in the plot. The perceived tempo seems to be somewhere between 143-155BPM, though this seems to fluctuate a bit in Spotify’s analysis. From these analyses we can see that Spotify also has a difficult time pinpointing the exact tempo played in the track.

Analyzing feature importance using Random Forest


Using a random forest model, I decided to measure the importance of each feature in an industrial hip-hop track. The results showed that energy and speechiness ranked among the most important features, along with timbre component 8 and the G key. One would expect energy and speechiness to be important features in any hip-hop track, though the G keys’ importance along with component 8 are quite interesting. Upon plotting these features, there seemed to be a slight correlation between component 8 and speechiness, with a high level of speechiness resulting in a higher level of component 8.

Conclusion

These set of analyses show that Industrial Hip-hop’s unique sound isn’t always an easy task for spotify to measure. While some tracks in this genre have a clear structure, others are not so obvious and difficult for Spotify to break down into its base components, as seen with a few of Death Grips’ and JPEGMAFIA’s tracks. Spotify in particular struggled with analyses of the tempo as well as the keys and chords identification as seen, for instance, in the keygram of JPEGMAFIA’s “HAZARD DUTY PAY!”.

From this we can perhaps derive that a Industrial Hip-Hop track is one which does not necessarily conform to the standard hip-hop structure. It shares similar elements, such as loudness and energy, yet it’s bizarre production and mixing often times leaves spotify confused and struggling, making it stand out from its other hip-hop counterparts.